AI: The Rise of the Machines... Or Just a Lot of Overhyped Chatbots?

Noodleface said:

I always found libre office barely passable and barely usable. I give Microsoft a lot of shit but their office products just work

Click to expand...

rhinohelix said:

That sub model that was printing billions, that they grew fat while everyone starved, <Christopher Walken King of New York reference>the genie is out of the bottle and will never go back in. The era of bespoke applications have arrived, and the Big Tech companies know, and knew it. I mean, lots people will keep paying, and there is value there: Those companies have spent millions over decades crafting and maintaining those apps, to be sure, but the idea that there are hostage monopolies who won't ever go anywhere else, and ideas like ribbons and buttons and icons are Nintendo-like patentable, are going to rapidly fade. It's at least part of the reason why the Big Tech are pouring billions/trillions into AI, because all their walled gardens become vulnerable, or at least assailable, in this new era, and they have to be able to offer these services as something that people are going to want to pay for, as other services fall off. It's 1995 and every brand is putting their URL in their commercials for the first time. It's not going to happen overnight, and not all at once but we are seeing those first exciting ripples of change.

Look at the OS Space: Omarchy in the Linux space is so configurable on the fly agentically; there will be hooks into Windows/an agentic version of Windows to be able to do this by Q3/4 next year. Microsoft clings to all of its desktop presence by being default but knows there is a real threat coming. Even though they make a bunch of money from other places, Azure/Datacenter/etc, M365/Office etc all depends on its defaultancy. They know their first AI efforts have failed and are desperately racing to catch up.

It's like secrets and encryption in any era of Quantum Computing: they almost become Whale Oil. From someone who could never code to someone who will have an app on the Apple App Store in a month, the scope of what's possible is rapidly changing, and regulation and the law/courts etc will never be able to keep up, just like when the Internet arrived.

Click to expand...

rhinohelix said:

That sub model that was printing billions, that they grew fat while everyone starved, <Christopher Walken King of New York reference>the genie is out of the bottle and will never go back in. The era of bespoke applications have arrived, and the Big Tech companies know, and knew it. I mean, lots people will keep paying, and there is value there: Those companies have spent millions over decades crafting and maintaining those apps, to be sure, but the idea that there are hostage monopolies who won't ever go anywhere else, and ideas like ribbons and buttons and icons are Nintendo-like patentable, are going to rapidly fade. It's at least part of the reason why the Big Tech are pouring billions/trillions into AI, because all their walled gardens become vulnerable, or at least assailable, in this new era, and they have to be able to offer these services as something that people are going to want to pay for, as other services fall off. It's 1995 and every brand is putting their URL in their commercials for the first time. It's not going to happen overnight, and not all at once but we are seeing those first exciting ripples of change.

Look at the OS Space: Omarchy in the Linux space is so configurable on the fly agentically; there will be hooks into Windows/an agentic version of Windows to be able to do this by Q3/4 next year. Microsoft clings to all of its desktop presence by being default but knows there is a real threat coming. Even though they make a bunch of money from other places, Azure/Datacenter/etc, M365/Office etc all depends on its defaultancy. They know their first AI efforts have failed and are desperately racing to catch up.

It's like secrets and encryption in any era of Quantum Computing: they almost become Whale Oil. From someone who could never code to someone who will have an app on the Apple App Store in a month, the scope of what's possible is rapidly changing, and regulation and the law/courts etc will never be able to keep up, just like when the Internet arrived.

Click to expand...

Noodleface said:

Depends how it's released.

If it's open source, you can just check the source and build it yourself. Pretty easy to sniff out illicit stuff in code (or have Claude scan it).

If it's closed source anything could be in there, just like anything you download.

Click to expand...

rhinohelix said:

Folks have already answered but generally its defense in depth:

1)Open-source is visible, allowing everyone to see what directly is in there to know what is in the code.
2) Git repos are generally available and people make it a point to scan them; also when you use/integrate/install them, your own AV should kick in and scan them.
3) Closed or wrapped software that's installed is signed by the people who give it to you, so there is an expected payload of what's in the file; if it doesn't match up with what's expected either end, if its not what its supposed to be, if something gets changed, all alarms bells go off. Typically its scanned on both ends before it gets signed, so you trust the signer and then you scan it yourself when you get it before you install it.

Now there are always edge cases and evil never sleeps but for the vast majority of normal use cases, that covers it. There have been exceptions; HWMonitor and CPUID download links were hijacked earlier this year: https://www.cyderes.com/howler-cell/how-cpuids-hwmonitor-supply-chain-was-hijacked-to-deploy-stx-rat. People found out pretty quickly, though, six hours later they already all over it.

So, the days of just going anywhere and downloading anything, Limewire, et al. are over for me, unless you are doing it in a virtual throwaway sandbox. Zero Trust isn't just branding. But for almost every well-travelled path, as long as you are protected on your end and cautious about what you do, you can be reasonably sure that someone would have stepped on the landmine first.

Targeted/broad phishing is a much greater threat, and is only getting more dangerous with AI smoothing out the usual tells.

Edit: I just really noticed the "False clean" part. I mentioned that there are edge cases and I don't know how far off the track we want to go here but generally in the multi-part assembly arms race it's hard for malicious code to hide. Typically (I am not Cybersecurity but work with them often) IIRC the most devious versions will wait until it can call home and download components. It's how they broke Solar Winds, IIRC. It's what LG TVs do now, per Gamers Nexus, record audio/video/info and then wait years to call home and upload it.

Click to expand...